Back to Founder Hub
✦ Founder Playbook

The Defensibility
Checklist

If you only rely on OpenAI's API, you have no moat. Here is how to build defensibility into your AI startup.

Data Flywheels
Custom Agents
Workflow Embedding

Investors are terrified of funding "Thin Wrappers"—startups that simply pass a user's prompt to ChatGPT and return the result. If OpenAI updates their UI tomorrow, your startup dies.

The Core Thesis: Your defensibility comes from what you do *before* and *after* the LLM call, not the LLM call itself.

The "Kill Zone" vs. The "Moat Zone"

Before diving into specific technical implementations, investors must understand where a product sits on the value spectrum:

1. The Proprietary Data Flywheel

The best AI startups use off-the-shelf models to acquire users, but use the data generated by those users to train proprietary fine-tuned models.

2. Custom Context Architecture (Complex RAG)

A basic RAG pipeline (chunking a PDF and doing vector search) is no longer a moat. Everyone knows how to do it.

3. High Switching Costs via Workflow Integration

If your product is just a chat box, the user can easily switch to ChatGPT. If your product is deeply embedded into their daily workflow, they can't.


4. Archetypes of Highly Defensible AI Startups

When evaluating early-stage companies, look for these structural patterns which indicate a deep moat:


5. The Investor Evaluation Framework

How VCs should measure the depth of an AI moat using real-world signals rather than hype.

Quantitative Indicators

Qualitative Indicators

5. Anti-Patterns & "Fake Moats"

Watch out for these massive red flags that indicate a startup is a "Thin Wrapper" residing entirely in the Kill Zone.